Statistical recursive filtering for offset nonuniformity estimation in infrared focal-plane-array sensors

We have developed a recursive estimator for the offset nonuniformity (NU) present in infrared (IR) focal-plane-array (FPA) imaging systems. The estimator is optimal in the least-square error sense and estimates the offset NU using only the collected scene data. Our scene-based nonuniformity correction (NUC) method exploits Kalman’s theory in order to estimate a constant in noise. The method is based upon two key assumptions: (i) the input irradiance at each detector is assumed to be a random variable uniformly distributed in a range that is common to all detectors in the IR–FPA; and (ii) the IR data sequences considered are temporally short enough such that the offset NU at each pixel of the array remains constant within the sequence. The proposed algorithm iterates over the frames in a pixel-by-pixel basis and updates two filter parameters for each pixel over the IR–FPA. Additionally, an on-line technique for ghosting artifacts cancellation is introduced to the NUC filter and an index to assess the artifact reduction is developed. The ability of the method to compensate for the offset NU is demonstrated by filtering video sequences from mid- and long-wave IR cameras.